Automating a Labour Performance Measurement and Risk
Assessment:
An Evaluation of Methods for a Computer Vision based
System
by
Donald Lloyd van Blommestein
$SULO
Thesis presented in fulfilment of the requirements for the degree of Master of Science in the Faculty of Industrial Engineering at
Stellenbosch University
i
Declaration
By submitting this thesis electronically, I declare that the entirety of the work contained therein is my own, original work, that I am the sole author thereof (unless to the extent explicitly otherwise stated), that reproduction and publication thereof by Stellenbosch University will not infringe any third party rights and that I have not previously in its entirety or in part submitted it for obtaining any qualification.
Date: ƉƌŝůϮϬϭϰ ŽƉLJƌŝŐŚƚΞϮϬϭϰ^ƚĞůůĞŶďŽƐĐŚhŶŝǀĞƌƐŝƚLJ ůůƌŝŐŚƚƐƌĞƐĞƌǀĞĚ
ii
Abstract
This thesis brings together productivity and risk assessments through innovative design, development and evaluation of a unique system for retrieving and analysing data. In the past, although the link between them is well-documented, these assessments have largely been dealt with as separate antagonist entities.
A broad evaluation of the existing traditional and technological support systems has been conducted to identify suitable methodologies along with a common technological platform for automation. The methodologies selected for the productivity and risk assessments were; work sampling and the revised NIOSH lifting equation respectively.
The automation of these procedures is facilitated through computer vision and the use of a range imaging Kinect™ camera. The standalone C++ application integrates two tracking approaches to extract real-time positional data on the worker and the work-piece. The OpenNI and OpenCV libraries are used to perform skeletal tracking and image recognition respectively. The skeletal tracker returns positional data on specific joints of the worker, while the image recognition component, a SURF implementation, is used to identify and track a specific work-piece within the capture frame. These tracking techniques are computationally expensive. In order to enable real time execution of the program, Nvidia’s CUDA toolkit and threading building blocks have been applied to reduce the processing time.
The performance measurement system is a continuous sampling derivative of work sampling. The speed of the worker’s hand movements and proximity to the work-piece are used to classify the worker in one of four possible states; busy, static, idle, or out of frame. In addition to the worker based performance measures, data relating to work-pieces are also calculated. These include the number of work-pieces processed by a specific worker, along with the average and variations in the processing times.
The risk assessment is an automated approach of the revised NIOSH lifting equation. The system calculates when a worker makes and/or breaks contact with the work-piece and uses the joint locations from the skeletal tracker to calculate the variables used in the determination of the multipliers and ultimately the recommended weight limit and lifting index. The final calculation indicates whether the worker is at risk of developing a musculoskeletal disorder. Additionally the information provided on each of the multipliers highlights which elements of the lifting task contribute the most to the risk. The user-interface design ensures that the system is easy to use. The interface also displays the results of the study enabling analysts to assess worker performance at any time in real time. The automated system therefore enables analysts to respond rapidly to rectify problems. The system also reduces the complexity of performing studies and it eliminates human errors. The time and costs required to perform the studies are reduced and the system can become a permanent fixture on factory floors. The development of the automated system opens the door for further development of the system to ultimately enable more detailed assessments of productivity and risk.
iii
Opsomming
Produktiwiteit en risiko evaluerings word in hierdie tesis saam hanteer deur die innoverende ontwerp, ontwikkeling en evaluering van 'n unieke stelsel vir die meting en ontleding van data. Alhoewel die skakel tussen hulle goed gedokumenteer is, word hierdie evaluering as afsonderlike antagonistiese entiteite hanteer.
'n Breë studie van die bestaande tradisionele en tegnologiese ondersteuningstelsels is gedoen om toepaslike metodes te identifiseer, om 'n gemeenskaplike tegnologiese platform vir outomatisering daar te stel. Die metodes wat gekies is vir die produktiwiteit en risiko bepalings is onderskeidelik werk monsterneming en die hersiende NIOSH opheffing vergelyking.
Die outomatisering van hierdie prosedures word gefasiliteer deur middel van rekenaar visie en die gebruik van 'n Kinect™ 3D kamera. Die selfstandige C++ program integreer ‘n dubbelvolgings benadering om in reële tyd posisionele data van die werker en die werk-stuk te kry. Die OpenNI en OpenCV biblioteke word onderskeidelik gebruik om skeletale volging en beeld erkenning uit te voer. Die skeletale volger bepaal posisionele data van spesifieke gewrigte van die werker, terwyl die beeld erkenning komponent, 'n SURF implementering gebruik om 'n spesifieke werk-stuk binne die opname raam te identifiseer en te volg. Hierdie volgings tegnieke is berekenings intensief. Om werklike tyd uitvoering van die program te verseker, is Nvidia se CUDA gereedskapstel en liggewig boublokke geimplementeer. Die produktiwiteit meting-stelsel is 'n aaneenlopende monsterneming benadering van werk monsterneming. Die spoed van die werker se handbewegings en nabyheid aan die werkstuk word gebruik om die werker te klassifiseer as in een van vier moontlike toestande; besig, staties, onaktief of buite die raam. Benewens die werker gebaseerde metings, word daar ook data oor werkstukke bereken. Dit sluit in die aantal werkstukke verwerk deur 'n spesifieke werker, sowel as die gemiddelde en variasie in verwerkings tye.
Die risiko-berekening is 'n outomatiese benadering van die hersiende NIOSH opheffing vergelyking. Die stelsel bereken wanneer die werker kontak maak en/of breek met die werkstuk en maak gebruik van die gewrigsposisies wat die skeletale volger aandui om die veranderlikes wat in die vermenigvuldigers gebruik word te bepaal. Die vermenigvuldigers word gebruik om die aanbevole maksimum gewig en die opheffing indeks te bereken. Die opheffing indeks dui aan of daar ‘n risiko vir die werker is om muskuloskeletale versteuring te ontwikkel. Benewens dui die vermenigvuldigers aan watter elemente die grootste bydra tot die risiko van die opheffingstaak maak.
Die gebruiker-koppelvlak-ontwerp verseker dat die stelsel maklik is om te gebruik. Die koppelvlak vertoon ook die resultate van die studie sodat ontleders op enige tyd werker prestasie kan evalueer in reële tyd. Die outomatiese stelsel stel dus ontleders in staat om vinnig te reageer sodat probleme reggestel kan word. Die stelsel verminder ook die kompleksiteit vir die uitvoering van studies en dit elimineer menslike foute. Die tyd en koste vereis om die studie te doen, word verminder en die stelsel kan ‘n permanente instelling op fabriekvloere geword. Die ontwikkeling van die outomatiese stelsel maak die deur oop vir verdere ontwikkeling van die stelsel om uiteindelik daartoe te lei dat meer gedetailleerde evaluering van produktiwiteit en risiko bepaal kan word.
iv
Acknowledgements
Firstly I would like to thank Dewald Swart for his time and assistance with the programming of the system developed in this thesis.
Then I would like to thank Dr Andre van der Merwe for his assistance and providing me with opportunities during the time of my studies.
I would also like to thank Mr Stephen Matope for his guidance and advice that assisted getting two journal articles published on a related topic.
Finally I would like to thank my parents and Retha Hamilton for their support, patience and understanding during my time of study.
v
Contents
Declaration ... i Abstract ... ii Opsomming ... iii Acknowledgements ... iv List of figures ... xiList of tables ... xiv
Nomenclature ... xvi
Chapter 1: Introduction Introduction to the report ... 2
1.1 Background ... 2 1.2 Aim ... 4 1.3 Problem statement ... 4 1.4 Justification ... 5 1.5 Delimitations ... 5 1.6 Limitations... 6 1.7 Definition of terms ... 6 1.8 Structure ... 9 1.9 Chapter 2: Literature study Introduction ... 12 2.1 Measurement methodologies ... 13 2.2 Traditional Analyses ... 14 2.3 Self-reports... 14 2.3.1 Manual observational techniques ... 14
2.3.2 Labour performance measurement ... 15
2.3.2.1 Risk assessment ... 18
2.3.2.2 Technological Support Systems ... 21
2.4 Labour performance measurement ... 21
2.4.1 Barriers impeding traditional work measurement ... 21
2.4.1.1 Analysis tools ... 23
2.4.1.2 Digital observational tools ... 25 2.4.1.3
vi
Comparison between two computer vision LPMs ... 33 2.4.2 Risk assessment ... 35 2.4.3 Analysis tools ... 35 2.4.3.1 Direct measurements ... 37 2.4.3.2
Digital observational techniques ... 39 2.4.3.3
Comparing techniques ... 43 2.5
Selection of a technological base and methodologies ... 45 2.6
Selecting a labour performance measurement system ... 45 2.6.1
Selecting a risk assessment methodology ... 46 2.6.2
Conclusion ... 46 2.7
Chapter 3: HAWK - the basic system
Concept of automation ... 48 3.1 The system ... 49 3.2 Hardware... 49 3.2.1 Software ... 54 3.2.2 Worker tracker ... 54 3.3 Selecting an SDK ... 55 3.3.1
The worker tracker application ... 56 3.3.2
Testing of the worker tracker ... 57 3.3.3
Motoman® SDA10 test ... 58 3.3.3.1 Human testing ... 63 3.3.3.2 Item tracker ... 70 3.4 SURF ... 72 3.4.1 OpenCV SURF_GPU ... 73 3.4.2
Random Sample and Consensus ... 75 3.4.3
Threading building blocks ... 77 3.4.3.1
Variables ... 78 3.4.3.2
Discussion ... 78 3.4.4
Item tracking inputs with the GUI ... 79 3.4.4.1
Operating the item tracker ... 80 3.4.4.2
Item tracker testing ... 84 3.4.5
vii
Processing speed and accuracy ... 85 3.4.5.1
Constants ... 86 3.4.5.2
Item size and distance from the camera... 86 3.4.5.3
Lighting ... 89 3.4.5.4
Selection modes ... 90 3.4.5.5
The tracking limit ... 91 3.4.5.6
Rotation ... 91 3.4.5.7
Contact between the worker and the work-piece ... 93 3.5
Background ... 95 3.5.1
Computing speed ... 96 3.5.2
Contact radius considerations ... 96 3.5.3
HAWK-RNLE contact considerations ... 97 3.5.3.1
HAWK-PRODUCTIVITY contact considerations ... 97 3.5.3.2
General contact considerations ... 97 3.5.3.3 Anthropometry ... 98 3.5.4 Type of grip ... 99 3.5.5 Accuracy evaluation ... 99 3.5.6
Data collection after contact... 101 3.5.7
Noise evaluation ... 103 3.5.8
Setting the contact radius ... 104 3.5.9
Conclusions on components ... 106 3.6
Chapter 4: The HAWK-PRODUCTIVITY system
Introduction ... 110 4.1
The concept of continuous sampling ... 110 4.2
Work sampling ... 111 4.3
Probability and classification theory ... 111 4.3.1 Performance indicators ... 112 4.3.2 Operator utilization ... 112 4.3.2.1 Allowances ... 112 4.3.2.2 Standard time ... 113 4.3.2.3 Automation ... 114 4.4
viii
Worker based measures ... 114 4.4.1 Productive ... 115 4.4.1.1 Static... 118 4.4.1.2 Idle... 119 4.4.1.3
Out of frame (Allowances) ... 120 4.4.1.4 Special cases ... 121 4.4.1.5 New measures ... 122 4.4.1.6 Conclusions ... 122 4.4.1.7
Work-piece based measures ... 122 4.4.2
The graphical user interface ... 123 4.5
Testing ... 126 4.6
Small item test ... 127 4.6.1
Design ... 128 4.6.1.1
Analysis ... 128 4.6.1.2
Discussion and conclusion ... 130 4.6.1.3
Large item test ... 131 4.6.2
Contact ... 131 4.6.2.1
Number of Items processed ... 132 4.6.2.2
The final system test ... 132 4.6.3
Design ... 133 4.6.3.1
Analysis ... 134 4.6.3.2
Discussion and conclusions ... 141 4.6.3.3
Conclusions ... 142 4.7
Results ... 143 4.7.1
Comparison to existing automated LPMs ... 145 4.7.2
Chapter 5: The HAWK-RNLE system
Introduction ... 148 5.1
Revised NIOSH lifting equation ... 148 5.2
Background ... 148 5.2.1
Revised NIOSH lifting equation formulation ... 149 5.2.2
Sensitivity analysis ... 152 5.2.3
ix
Measurement accuracy of human operators ... 153 5.2.4 Methodology ... 153 5.2.4.1 Findings ... 154 5.2.4.2 Automation ... 155 5.3 Existing system ... 155 5.3.1
The concept of automation... 156 5.3.2
The automation process ... 157 5.3.3
Methodology ... 160 5.3.4
The graphical user interface of results... 163 5.3.5 Problem areas ... 165 5.3.6 Testing ... 165 5.4 Preliminary test ... 166 5.4.1 Procedure ... 166 5.4.1.1
Findings of the preliminary test ... 166 5.4.1.2
Final testing procedure ... 167 5.4.2
Reference values ... 167 5.4.2.1
Considerations ... 169 5.4.2.2
Symmetric lifting task... 170 5.4.2.3
Asymmetric lifting task ... 172 5.4.2.4
Discussion of results ... 174 5.4.2.5
Frequency and coupling multipliers ... 177 5.4.2.6
Discussions and conclusions ... 179 5.5
Chapter 6: Discussions, conclusions and recommendations
Introduction ... 185 6.1
Summary and discussion ... 185 6.2
Findings ... 188 6.3
Concluding remarks ... 191 6.4
Future work and recommendations ... 192 6.5
Development of new measures ... 193 6.5.1
Development of a rotation invariant version ... 193 6.5.2
Technological upgrades ... 193 6.5.3
x References ... 195 Appendices Appendix 1 ... i Appendix 2 ... ii Appendix 3 ... v Appendix 4 ... vii Appendix 5 ... ix Appendix 6 ... xi Appendix 7 ...xiii Appendix 8 ... xiv Appendix 9 ... xv Appendix 10 ... xvii Appendix 11 ... xix Appendix 12 ... xx Appendix 13 ... xxii Appendix 14 ... xxiii Appendix 15 ... xxiv
xi
List of figures
Figure 1: Roadmap for the document ... 9
Figure 2: Methods for measuring labour performance and assessing risks ... 13
Figure 3: Technological Support Systems for Labour Performance Measurement ... 21
Figure 4: Screenshots of both the source and the filtered images. The green dot is located at the centre of the filtered object [13] ... 27
Figure 5: Screen shot of the computer screen at the factory [46] ... 28
Figure 6: Graph of the data recorded for a shift during the study [46] ... 29
Figure 7: Screenshot of training complete and bin locations fixed ... 31
Figure 8: (left) Image seen by the camera of test, (right) Screenshot of the performance feedback screen ... 32
Figure 9: Technological Support Systems for risk assessments ... 35
Figure 10: (Left) MVN BIOMECH 3D kinematics measurement system [56], (Right) Close up of optical markers [54] ... 38
Figure 11: Physical features of the BioHarnessTM BT by Zephyr [60] ... 39
Figure 12: Jack simulation of lifting a tyre [61] ... 40
Figure 13: Screenshot of Martin et al. Ergonomic Monitoring System [11] ... 43
Figure 14: Structured Light 3D Scanning by Measuring Distortion in Projected Stripe Patterns [12] ... 50
Figure 15: Microsoft Xbox 360 Kinect™ Camera and the corresponding co-ordinate system [71] ... 51
Figure 16: (Left) IR image of scene, (Centre) Distance image, (Right) Colour coded point cloud [73] ... 51
Figure 17: Internal Processing Layout of the Kinect™ [71] ... 52
Figure 18: Graphical User Interface for the worker tracker ... 56
Figure 19: Tracking the Motoman SDA10 ... 58
Figure 20: Motoman® SDA10 with end effectors in the tool mode ... 59
Figure 21: Motoman® tracking and alignment ... 60
Figure 22: Raw data worker tracker recorded ... 61
Figure 23: Worker tracker accuracy and noise test ... 65
Figure 24: Accuracy and noise testing on a worker’s feet ... 66
Figure 25: Noise of foot data used to identify a functional range ... 68
Figure 26: Number of correctly found matches per second for all of the conducted tests and algorithms [84] ... 71
Figure 27: Tracking a logo with OpenCV SURF_GPU implementation ... 74
Figure 28: Reduction in execution time enabled by TBB ... 77
Figure 29: Pin-hole camera model ... 79
Figure 30: GUI for SURF and RANSAC inputs ... 80
Figure 31: Selecting the tracking space ... 81
Figure 32: Selecting the work-piece ... 82
Figure 33: Curser location for selection for the work-piece ... 83
Figure 34: Item tracking, clearly tracking the selected work-piece ... 84
xii
Figure 36: Kinect™ depth sensor precision with distance. Measured values show quadratic relationship between the distance to the depth camera and the range and standard deviation of depth values [93] 88
Figure 37: Testing the impact on distance from the camera ... 88
Figure 38: Testing the impact on lighting on noise ... 89
Figure 39: (Left) Deficit selection, (Right) Credit selection ... 90
Figure 40: Tracking failure at 4.350m. (Left): Credit selection, (Right): Deficit selection ... 91
Figure 41: From left to right, Rotation testing around, x, y and z axis respectively ... 92
Figure 42: Conceptual representation of contact determination... 94
Figure 43: Tracking inaccuracy as contact is made with the work-piece... 96
Figure 44: A- Hook grip B- Power grip C- Ledge grip D- Palm pinch E- Finger or flat hand press ... 99
Figure 45: (Left) Contact accuracy determination for z dimension, (Right) Contact accuracy determination for x dimension ... 100
Figure 46: Measured and Actual distances ... 101
Figure 47: Measurement approaches ... 102
Figure 48: Photograph of the hand positions used in the contact test ... 103
Figure 49: Failure of the worker tracker in contact ... 104
Figure 50: User Inputs tab of the HAWK-PRDUCTIVITY GUI ... 124
Figure 51: Item monitors tab of the HAWK-PRDUCTIVITY GUI ... 125
Figure 52: Small item assembly test ... 127
Figure 53: (left) Worker in contact with the work-piece, (right) Worker stationary and not in contact with the work-piece ... 131
Figure 54: Item counter testing ... 132
Figure 55: (Left) Construction of the LEGO® bricks, (Right) Placing the LEGO® bricks through the corresponding holes and a fully constructed mesh drawn. ... 133
Figure 56: Timelines of the actual time spent in each state (all units in seconds) ... 136
Figure 57: Timeline of actual vs. measured values ... 139
Figure 58: Graphical representations of the measurement of task variables. (Left) Graphical representation of hand location, (right) Graphical representation of Angle of asymmetry [5] ... 151
Figure 59: Worker in the phi-pose required for calibration ... 157
Figure 60: The user inputs tab in the HAWK-RNLE GUI ... 158
Figure 61: Conceptual representation of the system ... 162
Figure 62: Results tab of the GUI ... 164
Figure 63: Screenshots of the symmetric lowering task. (Left) Origin on the stand, (Right) Destination on the floor ... 170
Figure 64: Screenshots of the asymmetric lifting task. (Left) Origin on table, (Right): Destination on stand ... 172
Figure 65: Screenshots of the GUI reflecting results of the frequency and coupling multiplier tests. (Top): Test 1, (Middle): Test 2, (Bottom): Test 3 ... 178
Figure 66: Demonstration of the worker trackers response to bending postures ... 182
Figure 67: Positional accuracy and tracking noise of workers hands ...xiii
xiii
Figure 69: Decision tree for coupling quality [5]... xv Figure 70: Screenshot of a text file populated with system times and coordinate data of a worker and a work piece ... xix Figure 71: Tables relating to the origin of a lift of the stand for the symmetric lifting test ... xx Figure 72: Tables for the assessment of the asymmetric lifting task at the origin of the lift from the table ... xxi Figure 73: HAWK-RNLE results tab for an uncompleted assessment not requiring significant control ... xxiii
xiv
List of tables
Table 1: Comparison between work sampling and time studies ... 18
Table 2: Exposure factors assessed by different methods [10] ... 19
Table 3: Results from the first shift of the case study ... 30
Table 4: Advantages of automated sampling ... 30
Table 5: Disadvantages of automated sampling ... 31
Table 6: Automated time study testing results... 32
Table 7: Advantages of the automated time study system ... 33
Table 8: Disadvantages of the automated time study system... 33
Table 9: Comparison between work sampling and time studies ... 34
Table 10: Comparison between manual and technological support systems ... 44
Table 11: Kinect™ hardware specifications [75] ... 53
Table 12: Accuracy and noise for the Motoman® SDA10 test ... 62
Table 13: Accuracy and noise for the foot accuracy test ... 67
Table 14: Item selection accuracy ... 86
Table 15: Constants for item tracker testing ... 86
Table 16: Results from testing on item size ... 87
Table 17: Distance test data ... 88
Table 18: Illumination level test ... 89
Table 19: Accuracy test results summary ... 100
Table 20: Contact radii ... 105
Table 21: Classification states criteria... 115
Table 22: Measurement errors of the small item assembly test ... 130
Table 23: Contact test results ... 131
Table 24: Accuracy of the average processing time determination ... 135
Table 25: Description and motivation behind the various classification states ... 135
Table 26: Complete system test data ... 137
Table 27: Altered Static and productive times for test two ... 138
Table 28: Classification criteria and sources for delays in the full system tests ... 142
Table 29: A comparison with existing techniques: The HAWK-PRODUCTIVITY system’s strengths and weaknesses. ... 145
Table 30: Criterion used to develop the revised NIOSH Lifting Equation [3]... 148
Table 31: Description of multipliers ... 149
Table 32: Equations, constraints and variable descriptions for RNLE multipliers [5,98] ... 150
Table 33: Range of possible values for variables and multipliers ... 152
Table 34: Calculations of variables in the HAWK-NIOSH system ... 161
Table 35: Steps in the proposed procedure to obtain the variables of the NIOSH Lifting Equation (NLE) [101] ... 168
Table 36: Symmetric lifting task results at the at the origin of the lift on the stand ... 171
xv
Table 38: Comparison between traditional study and the results of the HAWK-NIOSH system for a
symmetric lifting task [100] ... 172
Table 39: Asymmetric lifting task results at the at the origin on the lift on the table... 173
Table 40: Asymmetric lifting task results at the at the destination of the lift on the stand ... 173
Table 41: Comparison between traditional study and the results of the HAWK-NIOSH system for an asymmetric lifting task [100] ... 174
Table 42: Modified vertical distance with respective errors ... 175
Table 43: Reference and measured RWL for the different tests ... 180
Table 44: Advantages of the HAWK-PRODUCTIVITY system ... 192
Table 45; Disadvantages of the HAWK-PRODUCTIVITY system ... 192
Table 46: Coupling Multiplier [5] ... xv
Table 47: Frequency Multiplier ... xvi
Table 48: Summarized results from the measurements of a test lift (n = 27) [102] ... xvii
Table 49: Comparison between the reference and measured values for the five tasks [100] ... xviii
Table 50: Revised results for the origin of the symmetric lifting task with the proposed improvements implemented ... xxii
xvi
Nomenclature
AABB Axis Aligned Bounding Box
API Application Programming Interface
CMOS Complementary Metal–Oxide–Semiconductor CUDA Compute Unified Device Architecture
DALY Disability Adjusted Life Year GUI Graphical User Interface GPU Graphics Processing Unit HSE Health and Safety Executive
LI Lifting Index
LBP Low Back Pain
LPM Labour Performance Measurements
MODAPTS Modular Arrangement of Predetermined Time Standards MOST Maynard Operation Sequence Technique
MSD Musculoskeletal Disorder MTM Methods Time Measurement MVTA Multimedia Video Task Analysis
NIOSH National Institute for Occupational Safety and Health OCRA Concise Expose Index
OSHA Occupational Safety and Health Administration OWAS Ovako Working posture Analysis System PDA Personal Digital Assistant
PC Personal Computer
PMTS Predetermined Motion Time System PSM Physiological Status Monitoring RANSAC RANdom Sample and Consensus
xvii REBA Rapid Entire Body Assessment
RFID Radio Frequency Identification
RGB Red Green Blue
RNLE Revised NIOSH Lifting Equation RTLS Real-time Location Sensing RULA Rapid Upper Limb Assessment RWL Recommended Weight Limit SDK Software Development Kit
SIFT Scale-Invariant Feature Transform SOP Standard Operating Procedure SSP Static Strength Prediction SURF Speeded-Up Robust Features TBB Threading Building Blocks TSS Technological Support Systems USB Universal Serial Bus
UWB Ultra Wideband
Chapter 1
1.
2
Introduction to the report
1.1
This section provides the background for the project and highlights the importance and necessity of the project. It also states the aims and objectives of this project and defines the boundaries within which the project was undertaken. The key terms in the project description are also defined.
Background
1.2
The competitive and dynamic climate of global markets places significant pressure on companies. If companies are to remain profitable they need to continuously improve their service delivery. Effectively this entails producing higher quality products, quicker and at a lower cost to customers. The need to continuously improve and remain competitive is complicated further by technological developments. These developments frequently bring about changes; including updates and improvements to software, hardware and labour components in the production lines. In order to assess these improvements, various studies relating to the workforce and machinery need to be conducted.
In many developing countries automation is mostly restricted to large scale enterprises. The result is that many companies are still heavily dependent on labour. If companies are to remain profitable and competitive, the workforce needs to be highly productive. In many cases the performance of the workforce can be improved but for this to happen it firstly needs to be measured.
Numerous approaches for measuring labour performance have been developed. These methodologies have been extensively documented and utilised in production sectors around the world. The evaluation of labour performance typically identifies problem areas in the production process such as the under and overutilization of workers. The measurements also provide information relating to the manufacturing time along with the manufacturing costs of items. They also enable the determination of capacity. The basic function of performance measurements is to identify and evaluate possible changes so that productivity on factory floors is improved, where productivity is broadly defined as; Measurement of the efficiency at which components in the system convert inputs into outputs.
These traditional methods are supervisor intensive and have not developed much in concept since the introduction of work sampling in 1935. Traditionally the labour performance measurement has been done manually, using clerks or analysts to record observations. This traditional method is both time consuming, tedious and expensive. The techniques are also fairly intricate and require training in the various methods. The result is that the cost of performing analyses are high and subsequently many smaller companies only utilise measurements to a very limited degree or fail to perform studies at all. The driving force to increase productivity and maximize work-force utilization places strain on the workers. The excess pressure results in overexertion and work related musculoskeletal disorders (WMSD), such as Low Back Pain (LBP). These incidents form a significant financial burden in terms of direct costs such as compensation payments and in terms of indirect costs which originate from the reduction in productivity and disability [1,2]. In most industrialized countries, the costs of compensation for WMSDs account for at least one half of all workers compensation costs [1].
3
Approximately 70% of the global population suffers from LBP at some time in their lives [3,4]. Worldwide, in 2001, 37% of this fraction was attributed to occupational risk factors. The fraction varied between 21%-41% among regions and was higher in developing countries. LBP is one of the more common and most costly types of work injuries [2]. In the United States of America it amounted to 25% of all compensation payments and was the most costly [5]. Furthermore, it was the second most frequent reason for visits to the physician, the fifth-ranking cause of admission to hospital, and the third most common reason for surgical procedures [6]. The combined effect of occupational stressors were estimated to cause 818,000 Disability Adjusted Life years (DALYs) lost annually to LBP [4].
The National Research Council identified four work-related risk factors that show consistent and positive associations with the occurrence of LBP. These risk factors include [7]:
1. Manual materials handling 2. Frequent bending and twisting 3. Heavy physical load
4. Exposure to whole body vibration
Subsequent research has also found a positive association between other factors and LBP including [5,1]:
5. Awkward or extreme postures 6. Static postures
7. High once-off or accumulated forces on the spine
Over and above these physical causalities, psychological factors, such as stress and lifestyle factors have also been evaluated. Research has indicated that a positive correlation exists between poor social support in the workplace, low job satisfaction and the reporting of LBP. However, psychological factors in private life such as family support have not shown a positive correlation to the recording of back pain [6].
A substantial global effort has been made to reduce the occurrences of LBP. National and international laws and standards, such as ISO 11228-1: 2003 prescribe the methods for risk assessment associated with manual material handling [8]. These tools and systems pre-empt LBP cases by addressing the identified risk factors. A myriad of traditional and technological methods have been developed specifically to reduce the prevalence of LBP cases. The traditional methods are supervisor intensive, making them time consuming and tedious to perform. As technology has become available new methodologies and systems have been developed that aid and simplify the analysis of work elements. Many of the traditional and technological support systems that have been developed will be covered in this thesis.
The link between the labour performance measurements and risk assessments is well documented, yet the assessments are dealt with as separate antagonist entities. A number of technological developments with similar bases have been applied to both assessments. The wealth of technology narrows the gap between the assessments, indicating that the combination of the assessments is a distinct possibility.
4
A number of developments and improvements have been made to the traditional measurements techniques. Of particular interest is the non-invasive means of gathering data, such as the use of interactive video footage systems. These systems have ensured an improvement in the accuracy of performing labour performance measurements. [9] Similar technology has also been applied to assess WMSDs [10]. The biggest limitation of this approach is that the equipment and software required for the video analysis is rather expensive and as a result have not been implemented extensively.
The concept of using a video camera has merit. During the past few decades the computer scene has seen an exponential increase in the capacity and capability of computers. Furthermore the technological developments in cameras, particularly in range imaging cameras, enable the low cost image processing depiction of three-dimensional scenes. The reduced costs of hardware have also resulted in booming open source communities. These developments have dramatically reduced the costs and complexities of image processing. Real time image processing has already been applied in both performance measurements and risk assessment fields [11,12,13]
The wealth of available technology poses an excellent opportunity for the innovative development of a basic system that with minimal cost and effort can be used to evaluate labour performance and risk within in a workplace.
Aim
1.3
The aim of this thesis is to design and develop a system that is capable of autonomously and non-invasively gauging labour performance while also reducing the prevalence of musculoskeletal disorders.
Problem statement
1.4
This thesis was undertaken with the objective of automating both a traditional labour performance measurement and risk assessment. This requires the selection of a suitable technological platform and the assessment of the platform’s ability in the context of the proposed automation procedures. The automation includes both the data gathering and analysis phases of the traditional measurements. Additional objectives include improving the traditional approaches by reducing the time, costs and complexity associated with evaluating labour assessments. Ultimately, these improvements would ensure that the assessments would be available to a wider audience. The proposed systems should address the limitations of the exiting automated systems.
The data and results provided by the automated system should offer analysts an opportunity to improve the productivity of most organisations within the production sector. The results should also serve as reference for eliminating or reducing the risks that certain lifting tasks pose on workers.
5
Justification
1.5
Analysts in industry spend a significant amount of time gathering and analysing data relating to the workforce. The introduction of commercially available technological support systems has in many cases simplified the data gathering and/or analysis process. The vast majority of these systems use intrusive means of capturing data while the majority of non-intrusive devices function off-line.
There are very few commercially available tools that enable real time and non-intrusive monitoring of workers. The only real time, non-intrusive data gathering and/or analysis systems are based on computer vision. These systems enable the extraction of positional data of workers in real time by using live feeds from cameras linked to computers. The most recent development of low cost 3D range imaging cameras has resulted in rapid growth and development of open source image recognition libraries.
The availability of technological developments, such as computer vision enables the automation of the labour assessment methodologies proposed by this research.
Delimitations
1.6
The literature study does not cover all the techniques. Regarding Labour Performance Measurements (LPM) it is limited to those techniques requiring the identification of larger movements, and therefore excludes an evaluation of derived techniques such as predetermined time systems. The study also does not cover all topics relating to musculoskeletal disorders, but is limited to those that evaluate the prevalence of LBP.
The influence of ergonomic factors within the workplace are not considered, these include but are not limited to noise, lighting and temperature. The influences of these factors are considerable with regard to productivity and allowances. As a result they need to be dealt with separately.
The system that is developed in this thesis draws on existing methods and libraries. The functionality of these methods are tested within the context of this thesis, however the fundamental basis is not evaluated. It is assumed the basis and prior testing of the applied methods is scientifically sound. Throughout the development and write-up of this thesis, new libraries and software packages have been launched. Due to the rapid development occurring in the computer vision field, it was detrimental to progress of the thesis to keep updating to the latest software releases. This thesis was therefore conducted with fixed software versions, which at the conclusion of the thesis was already out-dated. The same applies to new hardware. New cameras have become available over the course of the thesis. The hardware available at initiation of the thesis was the hardware used throughout the thesis.
A vital component of the thesis is the determination of contact between the worker and the work-piece. Due to the nature of the image recognition algorithm as well as the programming complexity, contact is determined with an axis aligned approach. This approach requires approximate alignment between the work-piece and the camera’s co-ordinate system.
6
The system has been created to only track a single work-piece and a single worker within the capture frame at a time. Although the system has been created to be as stable as possible, errors relating to incorrect use have not been considered.
Worker performance ratings used in the calculation of standard times have been excluded from consideration. The average processing times calculated by the system are therefore not equal to the standard times.
Limitations
1.7
The thesis is constrained by both time and budget. The time constraints are 1800 working hours and the budget was limited to R70 000.
Definition of terms
1.8
Numerous terms used in this thesis have been explicitly defined here in order to prevent confusion as well as to provide a richer understanding of the text.
Effective
The system is effective if it is capable of fulfilling its purpose and therefore ultimately produces the desired results [14].
System
A system is an organized and co-ordinated method or procedure that has been formulated to accomplish a specific task. In the context of this thesis, the system refers to the group of interrelated and interdependent hardware and software components that are functionally connected and grouped together in order to execute the required analyses [14].
Productivity
It is defined as the measure of efficiency with which an activity converts inputs into value added outputs. Productivity is a relative measure. As a result the values themselves have little meaning. The values need to be compared with one another in order to be used [15,16].
Labour performance measurement (LPM)
Performance measurement is defined as the assessment of how well organizations are managed and the value of the services delivered [17]. In the context of this thesis LPM refers to the determination and evaluation of labour utilization, standard times as well as variations in working speed. Therefore how well workers are managed. The usefulness of LPM is determined by its ability to facilitate the improvement of productivity in the organization.
7
Risk Assessment
Risk is any event in which the outcome is uncertain and something of human value is at stake [18]. Risk is therefore a function of the likelihood and probability of an occurrence. In this thesis the term risk assessment is a contracted version of a longer phase and it is used specifically to describe the group of systems and techniques that quantify the risks of developing work related musculoskeletal disorders.
Labour assessments
The term labour assessment has been coined to simplify referring to both the LPM and the risk assessment.
Work related musculoskeletal disorder (WMSD)
Musculoskeletal disorders (MSD) refer to conditions that involve the nerves, tendons, muscles, and supporting structures of the body [19]. WMSDs, refers to the subset of MSDs that can be attributed to occupational stressors, including physical, psychosocial and individual risk factors [10] The specific area of interest in this thesis is the impact of exposure to physical factors in the workplace on the upper extremity and the lower back.
Traditional measurement
These are manually performed studies. They are typically the original concept.
Analysts
Refers to any individual that performs an assessment on workers.
Technological support system (TSS)
These are tools, devices and systems that have been created to simplify or automate labour assessments. They are not always linked to traditional measurements and may be unique systems.
Online (Real-time)
Refers to systems which guarantee responses within strict time constraints [20]. These systems are therefore capable of gathering and manipulating data before returning the results to the user within a set time from the moment they occur. In the context of this thesis the limit would be a few seconds.
Throughout the testing section numerous similar terms are used. In order to minimize confusion, this small section defines the most important terms.
8
Error
All measurements, regardless of the device used and type of measurement are only approximations. The deviations are referred to as errors. In an attempt to provide a distinction between the values of interest, wherever relevant other terms have been used.
Accuracy
It is the closeness with which the measurement of an element matches the true or actual value. It is calculated according to Equation 1. Accuracy takes systematic errors, random errors and resolution into account. The accepted value is defined by the measurement of the parameter with a system of known accuracy and precision.
| ̅ | … (1)
Noise
It represents the closeness with which repeated measurements of an element returns the same results. Due to the nature of this system, the term noise is substituted for precision. This is a result of the jumpiness of the data retrieved from the Kinect™ sensor. Noise per definition represents any unwanted addition to a signal. Two types of noise are presented:
Average noise (Noise) is calculated as the standard deviation. It is calculated according to Equation 2. Where E denotes the average of expected value of X.
√ [ ] … (2)
Maximum noise which provides information on the maximum discrepancy of the measured value. It is calculated according to Equation 3.
( ) ( ) … (3)
Systematic failures or errors
It refers to occurrences of factors which decrease the efficiency of the measurement system. These are consistent in that they are always prevalent under certain conditions [21].
Random errors
Are present in all systems and occur at random points in time. These are inherent in all systems and cannot be eliminated [21].
Relative Error
The ratio of the absolute error of the measurement to the accepted measurement, as indicated in Equation 4 The relative error expresses the "relative size of the error" of the measurement in relation to the measurement itself [21].
9
Structure
1.9
Figure 1: Roadmap for the document
Chapter 1 provides the backbone for the project by indicating its importance. It also highlighted the aims and objectives and defined the boundaries within which the project was undertaken. The key terms in the project’s description were also defined.
Chapter 2, as indicated in Figure 1, is separated into two primary categories; Labour performance measurement and risk assessments. These fields are further subdivided into the investigation of traditional methodologies and technological support systems. The concurrent investigation of both techniques is ultimately used to identify a common technological basis which will facilitate the automation of suitable traditional methodologies.
Chapter 3 introduces the concept for automation of the labour assessments. Following the basic concept, an investigation is conducted to identify hardware and software components with the desired characteristics. Chapter 3 covers the development of the basic computer vision based system, which can be broken down into three distinct components; worker-tracking, work piece tracking and contact determination. Each of these components is comprehensively tested to indicate the capabilities and limitations of the system.
Chapter 4 covers the automation of a LPM. The traditional work-sampling basis of the study, originally flagged for automation in Chapter 2, is discussed. Chapter 4 presents the process whereby the raw inputs from the basic system in Chapter 3 are processed to reveal information on the labour performance. The chapter concludes with testing on the system.
Chapter 5 covers the automation of the RNLE which was selected for automation of a risk assessment in Chapter 2. The traditional approach is comprehensively discussed and it provides the framework for the development of the automated version. Raw data from the basic system introduced in Chapter 3 is used to calculate the task variables of the RNLE. The results of tests done on the system are compared to the results of existing studies on human capabilities.
10
Chapter 6 provides a summary of the finding of the study. It also includes recommendations for future work.
11
CHAPTER 2
2.
12
Introduction
2.1
The aim of this section is to discuss, describe and compare the existing approaches and tools used to automate the measurement of labour performance and assess the risk of developing WMSDs. It is important to note that LPM and risk assessment are separate fields. They are brought together in this thesis because of their correlation and importance of considering them mutually.
There is constant and on-going effort being made to increase the productivity and maximize utilization of the workforce. Opportunities for improvement are often unveiled by gauging the current status of the workforce with LPMs. These measurements provide valuable information relating to standard times, rest allowances and utilization of workers. However, there is a risk that pressure to increase productivity could result in overexertion. There is also a risk that workers will adopt unsafe material handling practices in an attempt to save time. If workers develop WMSDs it could become a significant financial burden to companies. It is therefore important to ensure that tasks are safe and fall within the limits specified by the respective risk assessments.
One of the complexities is to establish the boundaries for a system that is capable of automating both labour performance measurements and risk assessment. The subsequent extensive literary research provides direction and guidelines for the selection of methodologies and hardware.
The literature can be divided into two distinct subtopics. Within both the LPM and risk assessment fields two approaches for executing studies exist:
Traditional measurements
Technological support systems
The traditional measurements use observers or analysts to gather and analyse data manually while technological support systems make use of technological advances to simplify or improve the data gathering and/or data analysis phases of the measurements.
The investigation of existing traditional LPMs and risk assessment methodologies provides valuable insight into the important characteristics that would need to be measured in the automated system. In addition understanding of the foundations of these measurements is vital in assessing their suitability for automation.
Technological developments have resulted in the creation of numerous systems that support the measurement and analysis phases of these measurements. A few fully automated systems have also been developed. The research of these systems brings to light the advantages, disadvantages and limitations relating to the use of each of these systems.
By considering the objectives of the project along with the evaluation of the traditional and technological support systems, a framework for the development of an effective and fully automated system is created.
13
Measurement methodologies
2.2
Numerous measurement methodologies have been developed to measure labour performance and to assess the risks of developing WMSDs. Many of the methodologies that have been developed for these assessments have similarities. Similar systems have been grouped together in order to provide a coherent structure for the literature review. The different approaches are categorised under the groups represented in Figure 2.
Figure 2: Methods for measuring labour performance and assessing risks
Traditional measurements are observer based approaches and include self-reports and manual observation techniques. Technological support systems utilise specialised hardware and/or software to retrieve and/or analyse data relating to the workers. These methods can be further classified as analyses tools, digital observational methods or direct measurements.
The analyses tools section includes systems and programs that have been specifically developed to reduce the complexities associated with the gathering and processing of data. Digital observational methods rely on the modification of digital inputs, to noninvasively extract data relating to workers. These approaches include computer vision or simulation based approaches. Direct measurements are only used to perform risk assessments. These systems utilize sensors or markers placed directly on the worker to extract data.
The various systems and approaches all have advantages and disadvantages, which along with the nuances between the assessment methodologies, are highlighted in the remainder of this chapter. The primary function of the methodologies is to enable companies to identify areas in which improvements can be made. In the context of this thesis the methodologies specifically enable decision makers to make
14
clear and concise decisions regarding the workforce. The methodologies also enable companies to gauge the impact of changes that have been implemented.
The effectiveness of the study is directly related to the inherent accuracy and relevance of the methodology. The time frame within which the analysis is conducted is equally important. The study will only be relevant if the time frame is truly representative of the working conditions. It follows that inaccurate or poorly timed measurements would result in warped decisions from management.
The time delay between the measurement and feedback of data is also important. Online and Offline systems have been developed and both approaches have merits. The advantage of the online approach is that problems and risks can be identified in real time. However, offline systems typically provide more and richer data. The trade-off between the two systems should be considered.
Traditional Analyses
2.3
Traditional analysis refers to observer based techniques. The observer is required to record measurements, perform calculations and analyse tasks and results. There are two categories: Self-reports and manual observation.
Self-reports
2.3.1
The data used in self-reports is generated and analysed by the worker. In labour performance measurements, self-assessments are often performed according to a sampling approach. The worker periodically records what he or she is doing at random predetermined times. A statistical analysis of the captured data provides insights into the time the worker is spending on various tasks [22].
Self-assessments are also used in the risk assessment field. In contrast to the statistical bases used in the LPM field, data is typically gathered from reports that workers make regarding physical and psychosocial factors. Other methods of data collection include interviews and questionnaires. Most recently, workers have been using video recordings of themselves and web based questionnaires [10].
Manual observational techniques
2.3.2
A variety of tools and methods have been developed that require an analyst to manually gather and analyse data. These tools typically form the basis of specific labour assessment fields as they represent the original concepts and methods [10]. A significant number of these approaches are widely used in industry and have been thoroughly scrutinised in academic circles. The investigations and assessment of approaches in this section will identify the most suitable candidates for automation.
In general the traditional measurements are typically practical for a wide range of applications. The biggest limitation is the influence of biases and inter-observer variability. These methods are also typically better suited for simple pattern tasks.
The basic execution of manual observational techniques starts well before the initiation of a study. Analysts are required to plan studies which also include preparatory work. For example, in LPMs this can entail determining the number of observations and the time at which measurements are to be taken.
15
Once the study is initiated the data is logged on task sheets and various calculations are performed on the data. The logging and analysis approaches will differ according to the assessment used. The data analysis phase can either be processed manually or with the aid of analysis tools. These tools have been specifically developed to simplify the processing of data and are widely available on the internet. One of the most common tools is an online calculator. The analyst simply inputs the recorded data into the specified cells and the calculator generates the results.
The following subsections provide insight into the manual observation techniques as used to asses LPM and risk assessments.
Labour performance measurement
2.3.2.1
There are numerous labour performance measurement techniques available to analysts. These include [23]:
1. Time and motion studies 2. Work sampling
3. Subjective evaluations 4. Reviewing records
5. Combinations of the techniques
Of these, only time studies and work sampling lend themselves to automation. These techniques will be discussed and compared in the following sub-sections.
Predetermined motion and time systems (PMTS) offer an alternative to the traditional labour performance measurements for establishing time standards. In contrast to time studies where the analyst times and subjectively rates the performance of an operator, PMTS requires the analyst to break a task down into its component actions. Time values are then assigned to each specific component action. Finally the analyst calculates the component times together to establish the standard time. There are a number of PMTS standards in existence today. The first PMTS was the methods-time measurement (MTM) released in 1948. Later Modular Arrangement of Predetermined Time Standards (MODAPTS) released in 1966, and Maynard Operation Sequence Technique (MOST) released in 1972, also gained popularity. At present, numerous variations of the MTM and MOST standards exist. These were developed to extend the range of application. Each of the various versions is suited to the assessment of specific tasks [22,24,25,26].
The PMTS methods require detailed analyses of the tasks. The methods therefore provide valuable detail on how the task is done. This enables the identification of possible areas for improvement. The methods also provide more accurate standard times. This results because the subjective evaluation of worker performance in work sampling and time studies is not required.
The automation of PMTS would certainly be valuable, and could be used to improve productivity. The methods are however limited to the establishment and optimization of production times. The objective
16
of this project is the real-time autonomous monitoring and assessment of the workforce. Due to the limited information provided by the methods, no further research was conducted.
Time and motion study
Time study is a methodology used to determine the standard time required to complete a task. The basic methodology involves [22,27,28]:
1. Selecting an average worker that has been trained in the specific task to be analysed. 2. Documenting the working conditions and other details relevant to the study.
3. Breaking the task down into subcomponents. The subcomponents are selected to be as small as possible without influencing the accuracy of the study. The analyst uses both sight and sound to determine when a subcomponent is started and ended.
4. Once the aforementioned steps are completed, the study is started using either a continuous or snapback timing method.
The calculated times are then deemed to be a fair representation of the time required to complete a given task, as it takes allowances and delays into consideration [22,27].
Time studies enable management to make effective decisions aimed at improving the efficiency of the entities operating within the system. It is important to realize that accurate time studies yield positive results and inaccurate time studies can create many problems [22,27].
Work sampling
Work sampling offers an alternative method for determining standard times and allowances. It therefore provides similar information to that acquired from time studies. However, it differs from time studies, in that it is an indirect measurement method [22,29].
The fundamental principle of work sampling is that it is based on the laws of probability. Work sampling was developed for the first time in 1935 by L.H.C. Tippet. It is the activity of taking randomly distributed observations of workers or machines with the primary objective of determining their utilization. This is achieved by classifying workers and machines as occupying one of a possible number of states at the time of the observation. The states are specified prior to the study.
The basic methodology involves [22,27]:
1. The objective of the study as well as the population of the study needs to be clearly defined. A list of the reasons why the study is being considered should be compiled.
2. The organization and its components to be studied need to be understood.
3. The different classification fields for the study need to be classified and documented. 4. Design and create a work sampling form that will be used in the study.
5. A preliminary estimate for critical factors needs to be determined. These can be derived from historical data or from a pilot study.
6. The desired accuracy of the study needs to be specified by the analyst. The analyst specifies two values that influence the accuracy of the study:
17
The confidence interval
The tolerance.
7. The estimate of the number of samples required for the work sampling study can then be determined.
8. Determine the frequency of the observations.
9. Folders for observers need to be prepared. Included in the folder should be the schedules, forms and layout plans.
10. The observers need to be trained.
A result of the statistical basis of work sampling is that the accuracy of the results is linked to the number of samples taken in the study. It is also influenced by the time period over which the assessment is performed [23]. The sampling basis of work studies enables analysts to observe multiple workers at the same time. As a result, work sampling can provide reasonably accurate representations of the conditions in the work environment at a fraction of the cost and without the drawbacks of continuous monitoring [29,22,23].
It is important for the analyst to understand the operations of the company so that an appropriate time frame for the study can be selected. Furthermore, the analyst needs to select a sample size that represents the true conditions of the system, while also understanding the capabilities of the observer. Some standards have been developed to aid in making selections about the number of samples in the study. One such guideline is to limit the number of observations to less than 8 per hour [23].
The objective of work sampling should be to attain unbiased results. These results should represent the true conditions of the individual under study. With the understanding that statistical methods form the basis of work sampling, the input data needs to be random and unbiased [29]. If these conditions are not met, the data will not deliver reasonable results. These inaccuracies can include [30]:
Continuity errors - Small changes in input data represent small changes in the output data.
Consistency errors - Similar runs will not reflect similar results. Comparison of traditional time studies and work sampling
From the comparison in Table 1, it is clear that both techniques have a number of advantages and disadvantages [29,22,23].
18
Table 1: Comparison between work sampling and time studies
Time studies Work Sampling
Easy to understand The statistical basis of the study may be difficult for workers and/or management to comprehend.
Requires extensive knowledge and understanding of the task Does not require an understanding of the task
Subject to both observer and worker biases The effect of biases is limited because it is only present when performance ratings are selected
Accurate timing figures are obtained Enough samples need to be made to ensure that the desired accuracy of the final results are achieved
One task analysed at a time and therefore labour intensive. Multiple work stations observed at the same time, therefor fewer observers required
Cyclic variations are not as well compensated for Observations typically made over an extended time period which decreases the effects of cyclic variations
Study is a continuous uninterrupted process and tedious to perform
The study can be interrupted at any time with a minimal effect on the results, and therefore not as tedious Establishes a systematic understanding of job tasks and
therefore facilitates the determination of the best methods which aid in the development of training programs.
Provides no specific information about the job
Hawthorn effect is likely to influence results Hawthorne effect is less likely due to the large number of observations made
Analyst does not move between workers It is not an economical solution if workers are spread out over a wide area
The two techniques also have a number of similarities. The systems play an important role in identifying labour costs. However, If multiple analysts are used the results could be invalid. Variations could result from inconstant ratings of worker performance. In work sampling there could also be additional variability in the level of detail recorded by analysts. For both methodologies the reliability of the results is dependent on the number of trails or observations taken by analysts.
The comparison has provided valuable insights into the advantages, disadvantages and limitations of each methodology. The single biggest difference between the methods is the complexity of cycle recognition algorithms that would need function reliably in the automation of time studies. However, this does not exclude it from consideration.
The selection of a methodology for automation is delayed until after the subsequent evaluation of the exiting technological support systems.
Risk assessment
2.3.2.2
A large number of risk assessment methodologies have been developed. The majority assess several critical physical exposure factors. Table 2 represents a comprehensive list of techniques and corresponding critical exposure factors as developed by David [10].
Since the development of revised NIOSH lifting equation in 1993 there has been a trend towards developing methodologies that consider the contribution of a number of the factors leading to WMSDs. The combined impact of these factors is represented in as a concise exposure index [31].
In Table 2 three candidate methodologies are highlighted. These methodologies were considered for selection due to their complete coverage of the exposure factors. It is important to note that the whole
19
body vibration, although a known cause of LBP does not form a serious risk in manufacturing environments. The vast majority of exposure originates from non-occupational sources such as use of motor vehicles [32]. It is therefore not considered in the selection of a methodology.
Table 2: Exposure factors assessed by different methods [10]
Technique Posture Load/Force Movement
frequency Duration Recovery Vibration Others
OWAS X X
Checklist X
RULA X X X
NIOSH Lifting Equation X X X X X X
PLIBEL X X X
The Strain Index X X X X X
OCRA X X X X X X X
QEC X X X X X X
Manual Handling
Guidance X X X X X X
REBA X X X X
FIOH Risk Factor Checklist X X X X X
ACGIH TLVs X X X X
LUBA X
Upper Limb Disorder
Guidance,HSG60 X X X X X X
MAC X X X X
The Revised NIOSH Lifting Equation
The Revised NIOSH Lifting Equation (RNLE) is widely used in industry to reduce the risk of LBP associated with manual lifting. Additionally it also reduces the possibility of shoulder and arm pain. It provides an empirical method for calculating the maximum mass of an object to be lifted. This mass represents the load that the vast majority of all healthy workers could lift over a substantial period of time, without an increased risk of developing lifting related LBP [3].
The load constant, which represents the maximum allowable load that can be safely lifted by the majority of the population under ideal conditions, is diminished by a number of factors. These factors include the location of the load, the coupling and frequency of lifts.
The fundamental inputs are positional and time data from a few key-points on the worker’s body and the work-piece. A number of existing technologies are capable of retrieving accurate positional data automatically. Furthermore, the calculations are also fairly straightforward. As a result the methodology lends itself well to automation.
The concise expose index
The concise expose index (OCRA index) is similar in concept to Revised NIOSH lifting equation [31]. The daily numbers of actions actually completed are compared with the corresponding number of recommended actions.
